Table 2.
Identification of CTL escape mutations using different techniquesa
| Mutation characteristics | Gene (HXB2 amino acid sites) of CTL epitope | HLA restriction (prevalence [π]) in Caucasians | Sample consensus amino acids of the CTL epitope (escape site underlined) | Inferred avg time (yr) to escape (1/ϕ) | Inferred avg time (yr) to reversion (1/ψ) | HLA association P values (q values) |
|
|---|---|---|---|---|---|---|---|
| Using a contingency table | With phylogenetic correction | ||||||
| Escape rapidly (<10 yr) and revert rapidly (<10 yr) | p24 gag (108–117) | B*57/58 (0.096) | TSTLQEQIGW | 0.5 | 3.4 | 1.6 × 10−6 (0.0014) | 0.0070 |
| nef (134–141) | A*24 (0.196) | RYPLTFGW | 0.9 | 5.0 | 2.6 × 10−5 (0.0074) | 0.033 | |
| p24 gag (174–184) | B*44 (0.211) | AEQASQEVKNW | 3.2 | 5.9 | 0.0015 (0.26) | 0.36 | |
| p17 gag (20–28) | A*03 (0.224) | RLRPGGKKK | 4.1 | 4.8 | 0.0078 (0.95) | 0.036 | |
| nef (84–92) | A*02 (0.438) | AALDLSHFL | 4.4 | 7.8 | 0.020 | 0.046 | |
| nef (90–97) | B*08 (0.143) | FLKEKGGL | 5.4 | 9.4 | 0.041 | 0.316 | |
| RT (392–401) | A*32 (0.077) | PIQKETWEAW | 6.4 | 2.4 | 0.0064 (0.92) | 0.0064 | |
| nef (68–76) | B*07 (0.166) | FPVRPQVPL | 7.2 | 4.0 | 0.072 | 0.025 | |
| p24 gag (131–140) | B*27 (0.073) | KRWIILGLNK | 8.7 | 9.9 | 0.038 | 0.010 | |
| nef (135–143) | B*18/53 (0.140) | YPLTFGWCY | 9.1 | 4.1 | 0.035 | 0.068 | |
| RT (156–166) | A*11 (0.133) | AIFQSSMTK | 9.4 | 5.6 | 0.086 | 0.031 | |
| Escape slightly more slowly (10–15 yr) and revert rapidly (<10 yr) | RT (128–135) | B*51 (0.129) | TAFTIPSI | 12 | 0.0 | 0.019 | 1.0 |
| nef (74–81) | B*35 (0.196) | VPLRPMTY | 12 | 3.3 | 0.089 | 0.089 | |
| RT (137–146) | B*18 (0.122) | NETPGIRYQY | 13 | 2.6 | 0.052 | 0.052 | |
| p24 gag (84–92) | B*07 (0.166) | HPVHAGPIA | 13 | 1.7 | 0.044 | 0.044 | |
| RT (244–252) | B*57 (0.057) | IVLPEKDSW | 14 | 0.1 | 0.093 | 0.093 | |
| RT (173–181) | A*30 (0.067) | KQNPDIVIY | 14 | 0.0 | 0.091 | 0.091 | |
| p24 gag (197–205) | B*08 (0.143) | DCKTILKAL | 15 | 2.2 | 0.056 | 0.056 | |
| HLA association P value of <0.05 using either technique | p24 gag (131–140) | B*27 (0.073) | KRWIILGLNK | 3.8 | 16 | 0.012 | 0.013 |
| nef (116–124) | B*57 (0.057) | HTQGYFPDW | 0.0 | 18 | 5.0 × 10−4 (0.11) | 0.0012 | |
| RT (128–135) | B*51 (0.129) | TAFTIPSI | 0.9 | 19 | 0.0094 | 0.015 | |
| nef (128–137) | B*07 (0.166) | TPGPGIRYPL | 0.0 | 22 | 0.046 | 0.077 | |
This table shows how our method for estimating escape and reversion rates from cross-sectional data can be used alongside HLA association studies to give a broader overview of how to classify epitopes of potential benefit in a CTL-inducing vaccine. For this analysis, multiple comparisons were made between sites in optimally defined CTL epitopes in gag, RT, and nef and their restricting HLAs. For each comparison (n = 862), the strength of the association between the HLA and the mutation at the site was measured using a standard contingency table approach and an approach that corrects for phylogenies. Fisher's exact P values are provided and highlighted (boldface) where less than 0.05. As a measure of association strength under multiple comparison correction, q values estimated using the Benjamini and Hochberg method are provided where available. These can be used as a rough guide for the relationship between q values and P values using this type of data. Two escape mutations (underlined) remained significant at a value of 0.05 after correction for multiple comparisons using either the Bonferroni correction (P < 5.8 × 10−5 = 0.05/862) or false-discovery rate control (q value < 0.05). For each comparison, escape rates and reversion rates were inferred by fitting the escape prevalence data to the model under the assumption of exponential growth of the epidemic, making use of equations S1 and S2 in the supplemental material. Mutations—and their corresponding CTL epitopes—were listed if they escape rapidly (average time to escape < 10 years) and revert rapidly (<10 years). Mutations are also listed if they escape slightly more slowly (10 to 15 years) but revert rapidly (<10 years). Finally, additional mutations with an HLA association P value of less than 0.05 using either method are listed. More details of the methods and model parameters used for the analysis are provided in Materials and Methods. Note that mutation away from threonine (T; underlined) in TSTLQEQIGW was also found to be strongly associated with B*57 alone. A standard contingency table approach returns a P value of 1.5 × 10−5 and a q value of 0.0065. With phylogenetic correction, the P value is 0.0069, but no q value is returned using the Benjamini and Hochberg method.